Leader: Giuliano Resce (UNIMOL); Other collaborator(s): UNIBO/UNIMIB/ISTAT
This task evaluates the socio-economic and environmental sustainability of care provision and proposes a Machine Learning (ML) model for allocating health funds among heterogeneous regions. The activities consist of two main stages: 1. identification of factors contributing to health needs and costs; 2 estimation of a novel allocation function. Data on health expenditure and its determinants will be retrieved by official statistics and administrative repositories and WP1. ML models will be implemented to leverage the wide range of sources used here. The implementation of a statistical/informatics system for monitoring, mapping, and forecasting health expenditure with the aim of providing guidance to the NHS for more efficient, effective, and equitable management of the increasing demand for healthcare of an ageing population will be designed and tested.
Brief description of the activities and of the intermediate results:
Publication of the first results of our work through a working paper (Economics & Statistics Discussion Papers, University of Molise) with the title “The Determinants of Missed Funding: Predicting the Paradox of Increased Need and Reduced Allocation”.
The work has been presented during the monthly reunion of the research group GRAPE of the Italian National Research Council on December 15, 2023.
The work has been presented at the 28th annual conference of the Italian Health Economics Association held in Rome on December 5-6, 2023.
The work benefited greatly from the feedback obtained during these events. Therefore, it is in a continuous phase of enhancement and future development.
We completed the predictive analysis of the Italian healthcare expenditure at regional level using ML algorithms. From an extensive literature review we identified the major drivers of the expenditure, then we constructed a novel database using official statistics at regional level over the period 2015-2019. Results are highly predictive. Furthermore, we were able to identify the core drivers of healthcare expenditure by studying the feature importance. Eventually, we disentangled the relationship between features and expense by constructing partial dependence plots.
Main policy, industrial and scientific implications
Accurate forecasting of healthcare costs is essential for making decisions, shaping policies, preparing finances, and managing resources effectively, but traditional econometric models fall short in addressing this policy challenge adequately. This paper introduces machine learning to predict healthcare expenditure in systems with heterogeneous regional needs. The Italian NHS is used as a case study, with administrative data spanning the years 1994 to 2019. The empirical analysis utilizes four machine learning algorithms (ElasticNet, Gradient Boosting, Random Forest, and Support Vector Regression) and a multivariate regression as a baseline. Gradient Boosting emerges as the superior algorithm in out-of-the-sample prediction performances; even when applied to 2019 data, the models trained up to 2018 demonstrate robust forecasting abilities. Important predictors of expenditure include temporal factors and technological progress, average family size and share of public expenditure over the total, regional area, population and share of foreign residents, GDP per capita and labor activity, and share of elderly population (75 years old and over). The remarkable effectiveness of the model demonstrates that machine learning can be efficiently employed to distribute national healthcare funds to areas with heterogeneous needs.
Brief description of the activities and of the intermediate results
The paper named “Enhancing Healthcare Cost Forecasting: A Machine Learning Model for Resource Allocation in Heterogeneous Regions” has been improved with a slight change in the title which now is the following: “The determinants of health expenditure: A machine learning approach “. It is currently under review.
Starting from this analysis we performed ad additional predictive Machine Learning (ML) predictive models with a reduced number of features. We decided to use only variables for which the data was also available at municipal level. Once obtained a good predictive mode, we use them to predict the expenditure at municipal level for year 2019. Therefore, we created a choropleth map which identifies what we could call “healthcare needs” at municipal level.
The work has been presented at the annual conference of the Italian Association of Regional Sciences held in Turin on September 4-6, 2024.
The work has been presented at the Spoke 5 meeting titled “Sostenibilità dei sistemi di cura dell’anziano in una società che invecchia” held in Termoli on September 11-13, 2024.
We performed a predictive analysis of local taxation decisions using Machine Learning algorithms. The analysis aimed at identifying whether it could be possible to predict the probability for a local policymaker to set the local surcharge on personal income tax to its maximum rate by leveraging on the fact that this municipal tax is “less visible” by taxpayers. Among the most important predictors emerges also how local policymakers tend to have a higher probability to set the rate to its maximum in municipality with higher elderly population. The resulting working paper is titled “Is Local Taxation Predictable? A Machine Learning Approach”. It is currently under review.
The work has been presented at the Italian Association of Public Economics held in Cagliari on September 12-13, 2024.
Writing of a policy brief which summarized the work conducted so far.
1. Brief description of the activities and of the intermediate results
Review of the working paper “Is Local Taxation Predictable? A Machine Learning Approach” based on academic feedback received.
Finalization of the policy brief. Here below the major recommendations are showed.
Recommendations
1. Integrate Machine Learning into Healthcare Forecasting
2. Enhance Data Collection and Accessibility
3. Address Regional Disparities in Healthcare Funding
4. Prepare for Demographic Changes
5. Reassess Funding Structures
Brief description of the activities and of the intermediate results
This analysis examines welfare expenditure trends across Italian municipalities, with a specific focus on elderly-related expenditure. Key findings from the 2014–2020 period include:
Proposed Next Steps for Analysis
To build upon these findings, we propose two potential directions for further research:
Prediction Model for Municipal Welfare Expenditure:
Develop a predictive model using advanced statistical or machine learning techniques to estimate welfare expenditure at the municipal level.
Policy Impact Assessment:
Assess the effects of specific state-level policies on municipal welfare expenditure.
This would involve identifying relevant policies implemented during the 2014–2020 period and analyzing their impact on municipalities’ expenditure patterns.